ISSN (e): 2250 – 3005 || Volume, 08 || Issue,10|| October – 2018 || International Journal of Computational Engineering Research (IJCER) www.ijceronline.com Open Access Journal Page 49 QuantitativeAnalysis of Sobel-Feldman Edge Detector for Brain Tissue Segmentation in Single-Channel MR Image Ghanshyam D. Parmar, Heena B. Sorathiya Department of Biomedical Engineering, GEC Gandhinagar,Gandhinagar –382028,Gujarat, India. Corresponding Author:Ghanshyam D. Parmar --------------------------------------------------------------------------------------------------------------------------------------- Date of Submission: 15-10-2018 Date of acceptance: 31-10-2018 --------------------------------------------------------------------------------------------------------------------------------------- I. INTRODUCTION Magnetic resonance imaging (MRI) or nuclear magnetic resonance imaging (NMRI)[1], [2] is primarily medical imaging technique used in radiology to visualize internal structure of the body. MRI provides much greater contrast among different soft tissues of body. This ability makes it useful for neurological, musculoskeletal, cardiovascular and oncological imaging [3]. Brain matter could be generally categorized as White Matter (WM), Gray Matter (GM) and Cerebrospinal Fluid (CSF) [4], [5]. Most of brain structures are anatomically defined by the boundaries of these tissue classes[4]–[6]. So, we need a method of segmenting tissues in classes. It is an important step for quantitative analysis of the brain and its anatomical structures. Brain tissue classification is also an important step for detection of various pathological conditions affecting brains parenchyma[7]–[9]. It is also used for surgical planning and simulation [10]and three-dimensional visualization for diagnosis and detection of abnormalities[11]–[13]. It is also useful in the study of brain development [13]–[15] and human aging [15], [16]. In MR imaging, images are produced based on intensities achieved by three tissue characteristics namely: T1 relaxation time, T2 relaxation time and proton density (PD). The images obtained by these properties are known as T1- weighted MR images, T2-weighted MR images and proton density MR images respectively. The effect of these parameters image can be varied based on the adjusting the parameters like time to echo (TE) and time to repeat of the pulse sequence [17]. By using different parameters or number of echoes in the pulse sequence, a multitude of nearly registered images with different characteristics of same object can be achieved. If only a single MR image of the object is available such an image is referred to as single-channel (single-echo) image, and in case when number of MR images of the same object at same section are obtained, they are referred as multi-channel (multispectral or multi-echo) images[18]. For a given scanning time, the voxel sizes achieved in multi-spectral images are larger than those achieved with single-channel images. This ability of finer voxel sizes makes single-channel image more suitable for precise and accurate quantitative measurements of anatomical structures and tissues. Nevertheless, multichannel image provides more information at given voxel size than single-channel image[17], [18]. Most of segmentation techniques have relied on multi-spectral characteristics of ABSTRACT Segmentation of brain tissues is one important process prior to many analysis and visualization tasks for Magnetic Resonance (MR) images. Edge is one of the important characteristic features used in many image segmentation techniques for brain tissue segmentation in MR image. Sobel- Feldman approximation is technique used for detection of edges in any image. Unfortunately, MR images always contain significant amount of noise caused by operator performance, equipment and the environment. This noise can lead to major inaccuracies in edge detection process and hence in segmentation result. We conduct the research in measuring the performance of Sobel-Feldman approximation for edge detection in different noise level for single-channel MR image. To validate the accuracy and robustness of Sobel-Feldman approximation we carried out experiments on simulated MR brain scans. The performance of edge detector is analyzed by different quantitative measures. These quantitative measures include the mathematical measures like mean square error, signal to noise ratio and peak signal to noise ratio as well the statistical measures like accuracy, sensitivity, specificity and F measure. KEYWORDS: Brain tissue classification, Edge detection, F measure Magnetic Resonance, MR Images, Segmentation, Sensitivity,Sobel-Feldman approximation, Specificity